Transportation & Logistics​
Artificial Intelligence & Data

Boosting Revenue with AI-Powered Freight Matching: 18% Sales Growth for National Logistics Leader

A top-50 U.S. freight broker was losing revenue to manual load matching, high empty-mile rates, and pricing decisions made on gut instinct. RTS Labs deployed an AI-powered freight matching engine in 10 weeks — driving 18% revenue growth, cutting empty miles by 31%, and reducing match time from 47 minutes to under 4.

logistics supply chain header
Case Study at a Glance
Client

National Logistics Leader

Industry

Transportation & Logistics

Use Case

AI-Powered Freight Matching & Load Optimization

Tech Stack

Python

XGBoost

AWS SageMaker

Apache Kafka

Time to Production
From brief to live deployment
0 weeks

Dealing with a similar challenge?

Our engineers will map your workflow and define a ship date in a 2-week Discovery Sprint.

Share

1. The Challenge

The client, a top-50 U.S. freight broker by load volume, was growing fast — and their operations were starting to buckle under the weight of that growth. With over 1,200 active carrier relationships and thousands of load requests moving through their system each week, dispatchers were still making matching decisions the same way they had at 200 loads per day: spreadsheets, phone calls, and institutional knowledge.

The cracks were measurable. Empty miles — trucks completing a delivery with no return load — were running at 34%, well above the 22% industry benchmark. Match confirmation was taking an average of 47 minutes, meaning time-sensitive freight was routinely being offered to competitors who could respond faster. And because matching quality depended on which dispatcher handled the load, pricing and carrier selection varied significantly across the team — leaving both margin and customer experience on the table.

Empty Mile Drain

Trucks were completing 34% of miles without a paying load — well above the standard 22% industry benchmark — directly compressing the net profit margin on every route the company operated.

Empty miles vs. 22% benchmark
0 %

Match Latency

Dispatchers averaged an excessive 47 minutes to confirm a load-to-carrier match. Time-sensitive shipments were going to competitors who could respond in minutes, not nearly an hour.

Avg. time per match, manually
0 min

Pricing Left Behind

Lane pricing was set by dispatcher intuition instead of real-time market signals. On high-demand corridors, the company was systematically underpricing — surrendering margin on valuable freight.

Loads priced below market rate
0 %

2. The Engineer Approach

RTS Labs approached freight matching as a real-time recommendation problem, not a rules engine. Rather than replacing dispatchers, the system was designed to give them the right answer in seconds — surfacing ranked carrier recommendations with confidence scores and automated pricing guidance, and preserving human judgment for edge cases. The architecture centered on a gradient boosting model trained on 18 months of historical load and carrier data, served via a low-latency API integrated directly into the client’s existing TMS.

  • Data Integration & Feature Engineering

    Ingested 18 months of historical load, carrier, and lane data from the client's TMS and three external lane pricing APIs. Built a real-time event pipeline using Apache Kafka to stream incoming load requests and carrier availability state into a PostgreSQL feature store, with sub-second update latency.

  • ML Model Training & Lane Pricing

    Trained an XGBoost gradient boosting model on historical match outcomes to score carrier-load affinity by lane, time window, cargo type, and carrier on-time history. A separate regression model estimated prevailing market rate per lane using external pricing signals — giving dispatchers an automated pricing recommendation on every load.

  • Matching API & TMS Integration

    Built a REST API that accepts an incoming load and returns a ranked list of the top 5 carrier matches with confidence scores, estimated transit time, and suggested pricing. Integrated directly into the client's TMS via webhook — no UI change required for dispatchers. Match recommendations appear inline in their existing workflow.

  • Deployment & Observability

    Deployed inference endpoints on AWS SageMaker with auto-scaling to handle peak load windows. Built a real-time operations dashboard tracking match acceptance rate, empty-mile rate, pricing recommendation adoption, and model drift indicators — giving operations leadership a live view of AI performance and business impact.

Our operations were hitting a wall trying to scale manual load matching for thousands of weekly requests. RTS Labs stepped in and completely re-engineered the process in just 10 weeks. By transforming our siloed historical data into a real-time matching engine, they shifted our dispatchers from being operational bottlenecks to exception handlers. The results were immediate: cutting empty miles by 31% and dropping our match confirmation times from 47 minutes down to under 4. It’s been a massive win for our net profit margins and our capacity.
VP of Operations
Top-50 U.S. Freight Brokerage

3. Results & Impact

Revenue Growth
0 %
Faster Load Matching
0 x
Fewer Empty Miles
0 %
Time to Production
0 wks

Before RTS Labs

  • High Empty-Mile Drain

    34% empty-mile rate — well above the 22% industry benchmark — eroding margin on every route

  • Slow Manual Dispatch

    47-minute average match confirmation time, losing time-sensitive freight to faster competitors

  • Intuition-Based Pricing

    Lane pricing set by dispatcher intuition, with 12% of loads systematically underpriced on high-demand corridors

  • Fragmented Performance Data

    No visibility into match quality, pricing decisions, or carrier performance trends across the team

After RTS Labs

  • Maximized Fleet Capacity

    Empty-mile rate dropped to 23%, approaching industry benchmark and recovering meaningful margin per route

  • Accelerated Match Speed

    Average match confirmation time cut to under 4 minutes, with ranked carrier recommendations surfaced instantly

  • Data-Driven Pricing Engine

    Automated pricing recommendations on every load using real-time market-rate signals, with dispatcher override preserved

  • Real-Time Operational Insights

    Live operations dashboard tracking match acceptance, empty miles, pricing adoption, and model health in real time

More Case Studies:

Consulting Case Study

Transforming Executive 360 Analysis with AI – Schachter Consulting

Overview:

Case Study Finance​ Service Type Transforming Executive 360 Analysis with AI: 3 Days to 2 Hours for Schachter Consulting Schachter Consulting — a boutique firm

...
Financial analysts case study

Automating Financial Due Diligence with AI

Overview:

Case Study Financial Services & Due Diligence​ Service Type Automating High-Stakes Financial Due Diligence: 50% Faster Reports, 3x Client Capacity for Illume Illume’s financial analysts

...
Data Engineer Case Study

Modernizing a Legacy Database with AI for Stride

Overview:

Case Study Industry Service Type Modernizing a Legacy Database with AI: 80% Faster Queries, 4-Month Migration Done in 3 Weeks for Stride Stride’s legacy database

...
Have a Similar Challenge?

Every AI Idea Deserves a Ship Date

Stop strategizing. Start building. Let’s map your workflow and get your AI integration into production in 90 days.